]> git.proxmox.com Git - mirror_qemu.git/blame - block/parallels.c
block/snapshot: stress that we fallback to primary child
[mirror_qemu.git] / block / parallels.c
CommitLineData
6ada7453
TS
1/*
2 * Block driver for Parallels disk image format
3 *
4 * Copyright (c) 2007 Alex Beregszaszi
cc5690f2 5 * Copyright (c) 2015 Denis V. Lunev <den@openvz.org>
6ada7453 6 *
cc5690f2
DL
7 * This code was originally based on comparing different disk images created
8 * by Parallels. Currently it is based on opened OpenVZ sources
9 * available at
10 * http://git.openvz.org/?p=ploop;a=summary
6ada7453
TS
11 *
12 * Permission is hereby granted, free of charge, to any person obtaining a copy
13 * of this software and associated documentation files (the "Software"), to deal
14 * in the Software without restriction, including without limitation the rights
15 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
16 * copies of the Software, and to permit persons to whom the Software is
17 * furnished to do so, subject to the following conditions:
18 *
19 * The above copyright notice and this permission notice shall be included in
20 * all copies or substantial portions of the Software.
21 *
22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
23 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
24 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
25 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
26 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
27 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
28 * THE SOFTWARE.
29 */
922a01a0 30
80c71a24 31#include "qemu/osdep.h"
baefd977 32#include "qemu/error-report.h"
da34e65c 33#include "qapi/error.h"
737e150e 34#include "block/block_int.h"
609f45ea 35#include "block/qdict.h"
8942764f 36#include "sysemu/block-backend.h"
1de7afc9 37#include "qemu/module.h"
922a01a0 38#include "qemu/option.h"
1511b490
KW
39#include "qapi/qmp/qdict.h"
40#include "qapi/qobject-input-visitor.h"
41#include "qapi/qapi-visit-block-core.h"
58369e22 42#include "qemu/bswap.h"
0d31c7c2 43#include "qemu/bitmap.h"
5df022cf 44#include "qemu/memalign.h"
1d0f37cf 45#include "migration/blocker.h"
90fe66f0 46#include "parallels.h"
6ada7453
TS
47
48/**************************************************************/
49
50#define HEADER_MAGIC "WithoutFreeSpace"
d25d5980 51#define HEADER_MAGIC2 "WithouFreSpacExt"
6ada7453 52#define HEADER_VERSION 2
6dd6b9f1 53#define HEADER_INUSE_MAGIC (0x746F6E59)
555a608c 54#define MAX_PARALLELS_IMAGE_FACTOR (1ull << 32)
6ada7453 55
f7abe0ec
MAL
56static QEnumLookup prealloc_mode_lookup = {
57 .array = (const char *const[]) {
58 "falloc",
59 "truncate",
f7abe0ec
MAL
60 },
61 .size = PRL_PREALLOC_MODE__MAX
d6179011
DL
62};
63
d6179011
DL
64#define PARALLELS_OPT_PREALLOC_MODE "prealloc-mode"
65#define PARALLELS_OPT_PREALLOC_SIZE "prealloc-size"
66
67static QemuOptsList parallels_runtime_opts = {
68 .name = "parallels",
69 .head = QTAILQ_HEAD_INITIALIZER(parallels_runtime_opts.head),
70 .desc = {
71 {
72 .name = PARALLELS_OPT_PREALLOC_SIZE,
73 .type = QEMU_OPT_SIZE,
74 .help = "Preallocation size on image expansion",
ff5bbe56 75 .def_value_str = "128M",
d6179011
DL
76 },
77 {
78 .name = PARALLELS_OPT_PREALLOC_MODE,
79 .type = QEMU_OPT_STRING,
80 .help = "Preallocation mode on image expansion "
81 "(allowed values: falloc, truncate)",
82 .def_value_str = "falloc",
83 },
84 { /* end of list */ },
85 },
86};
87
1511b490
KW
88static QemuOptsList parallels_create_opts = {
89 .name = "parallels-create-opts",
90 .head = QTAILQ_HEAD_INITIALIZER(parallels_create_opts.head),
91 .desc = {
92 {
93 .name = BLOCK_OPT_SIZE,
94 .type = QEMU_OPT_SIZE,
95 .help = "Virtual disk size",
96 },
97 {
98 .name = BLOCK_OPT_CLUSTER_SIZE,
99 .type = QEMU_OPT_SIZE,
100 .help = "Parallels image cluster size",
101 .def_value_str = stringify(DEFAULT_CLUSTER_SIZE),
102 },
103 { /* end of list */ }
104 }
105};
106
d6179011 107
555cc9d9
DL
108static int64_t bat2sect(BDRVParallelsState *s, uint32_t idx)
109{
dd97cdc0 110 return (uint64_t)le32_to_cpu(s->bat_bitmap[idx]) * s->off_multiplier;
555cc9d9
DL
111}
112
2d68e22e
DL
113static uint32_t bat_entry_off(uint32_t idx)
114{
115 return sizeof(ParallelsHeader) + sizeof(uint32_t) * idx;
116}
117
29442569 118static int64_t seek_to_sector(BDRVParallelsState *s, int64_t sector_num)
6ada7453 119{
c34d2451 120 uint32_t index, offset;
6ada7453
TS
121
122 index = sector_num / s->tracks;
123 offset = sector_num % s->tracks;
124
9d8b88f6 125 /* not allocated */
369f7de9 126 if ((index >= s->bat_size) || (s->bat_bitmap[index] == 0)) {
f08e2f84 127 return -1;
369f7de9 128 }
555cc9d9 129 return bat2sect(s, index) + offset;
6ada7453
TS
130}
131
9de9da17
RK
132static int cluster_remainder(BDRVParallelsState *s, int64_t sector_num,
133 int nb_sectors)
134{
135 int ret = s->tracks - sector_num % s->tracks;
136 return MIN(nb_sectors, ret);
137}
138
6953d920
DL
139static int64_t block_status(BDRVParallelsState *s, int64_t sector_num,
140 int nb_sectors, int *pnum)
141{
142 int64_t start_off = -2, prev_end_off = -2;
143
144 *pnum = 0;
145 while (nb_sectors > 0 || start_off == -2) {
146 int64_t offset = seek_to_sector(s, sector_num);
147 int to_end;
148
149 if (start_off == -2) {
150 start_off = offset;
151 prev_end_off = offset;
152 } else if (offset != prev_end_off) {
153 break;
154 }
155
156 to_end = cluster_remainder(s, sector_num, nb_sectors);
157 nb_sectors -= to_end;
158 sector_num += to_end;
159 *pnum += to_end;
160
161 if (offset > 0) {
162 prev_end_off += to_end;
163 }
164 }
165 return start_off;
166}
167
7644d2dc
PB
168static coroutine_fn int64_t allocate_clusters(BlockDriverState *bs,
169 int64_t sector_num,
170 int nb_sectors, int *pnum)
5a41e1fa 171{
bda4cdcb 172 int ret = 0;
5a41e1fa 173 BDRVParallelsState *s = bs->opaque;
d8b83e37 174 int64_t pos, space, idx, to_allocate, i, len;
5a41e1fa 175
ddd2ef2c
DL
176 pos = block_status(s, sector_num, nb_sectors, pnum);
177 if (pos > 0) {
178 return pos;
179 }
5a41e1fa 180
ddd2ef2c 181 idx = sector_num / s->tracks;
969401fe 182 to_allocate = DIV_ROUND_UP(sector_num + *pnum, s->tracks) - idx;
86d1bd70
HR
183
184 /* This function is called only by parallels_co_writev(), which will never
185 * pass a sector_num at or beyond the end of the image (because the block
186 * layer never passes such a sector_num to that function). Therefore, idx
187 * is always below s->bat_size.
188 * block_status() will limit *pnum so that sector_num + *pnum will not
189 * exceed the image end. Therefore, idx + to_allocate cannot exceed
190 * s->bat_size.
191 * Note that s->bat_size is an unsigned int, therefore idx + to_allocate
192 * will always fit into a uint32_t. */
193 assert(idx < s->bat_size && idx + to_allocate <= s->bat_size);
194
ddd2ef2c 195 space = to_allocate * s->tracks;
d8b83e37
DL
196 len = bdrv_getlength(bs->file->bs);
197 if (len < 0) {
198 return len;
199 }
200 if (s->data_end + space > (len >> BDRV_SECTOR_BITS)) {
ddd2ef2c 201 space += s->prealloc_size;
bda4cdcb
EB
202 /*
203 * We require the expanded size to read back as zero. If the
204 * user permitted truncation, we try that; but if it fails, we
205 * force the safer-but-slower fallocate.
206 */
207 if (s->prealloc_mode == PRL_PREALLOC_MODE_TRUNCATE) {
208 ret = bdrv_truncate(bs->file,
209 (s->data_end + space) << BDRV_SECTOR_BITS,
210 false, PREALLOC_MODE_OFF, BDRV_REQ_ZERO_WRITE,
211 NULL);
212 if (ret == -ENOTSUP) {
213 s->prealloc_mode = PRL_PREALLOC_MODE_FALLOCATE;
214 }
215 }
19f5dc15 216 if (s->prealloc_mode == PRL_PREALLOC_MODE_FALLOCATE) {
720ff280 217 ret = bdrv_pwrite_zeroes(bs->file,
74021bc4
EB
218 s->data_end << BDRV_SECTOR_BITS,
219 space << BDRV_SECTOR_BITS, 0);
19f5dc15
DL
220 }
221 if (ret < 0) {
222 return ret;
223 }
5a41e1fa
DL
224 }
225
bcbb3866
EK
226 /* Try to read from backing to fill empty clusters
227 * FIXME: 1. previous write_zeroes may be redundant
228 * 2. most of data we read from backing will be rewritten by
229 * parallels_co_writev. On aligned-to-cluster write we do not need
230 * this read at all.
231 * 3. it would be good to combine write of data from backing and new
232 * data into one write call */
233 if (bs->backing) {
234 int64_t nb_cow_sectors = to_allocate * s->tracks;
235 int64_t nb_cow_bytes = nb_cow_sectors << BDRV_SECTOR_BITS;
a4072543 236 void *buf = qemu_blockalign(bs, nb_cow_bytes);
bcbb3866 237
a4072543
VSO
238 ret = bdrv_co_pread(bs->backing, idx * s->tracks * BDRV_SECTOR_SIZE,
239 nb_cow_bytes, buf, 0);
bcbb3866 240 if (ret < 0) {
a4072543 241 qemu_vfree(buf);
bcbb3866
EK
242 return ret;
243 }
244
eba088f9
HR
245 ret = bdrv_co_pwrite(bs->file, s->data_end * BDRV_SECTOR_SIZE,
246 nb_cow_bytes, buf, 0);
a4072543 247 qemu_vfree(buf);
bcbb3866
EK
248 if (ret < 0) {
249 return ret;
250 }
251 }
252
ddd2ef2c
DL
253 for (i = 0; i < to_allocate; i++) {
254 s->bat_bitmap[idx + i] = cpu_to_le32(s->data_end / s->off_multiplier);
255 s->data_end += s->tracks;
256 bitmap_set(s->bat_dirty_bmap,
c9f6856d 257 bat_entry_off(idx + i) / s->bat_dirty_block, 1);
ddd2ef2c 258 }
0d31c7c2 259
ddd2ef2c 260 return bat2sect(s, idx) + sector_num % s->tracks;
5a41e1fa
DL
261}
262
0d31c7c2
DL
263
264static coroutine_fn int parallels_co_flush_to_os(BlockDriverState *bs)
265{
266 BDRVParallelsState *s = bs->opaque;
267 unsigned long size = DIV_ROUND_UP(s->header_size, s->bat_dirty_block);
268 unsigned long bit;
269
270 qemu_co_mutex_lock(&s->lock);
271
272 bit = find_first_bit(s->bat_dirty_bmap, size);
273 while (bit < size) {
274 uint32_t off = bit * s->bat_dirty_block;
275 uint32_t to_write = s->bat_dirty_block;
276 int ret;
277
278 if (off + to_write > s->header_size) {
279 to_write = s->header_size - off;
280 }
32cc71de 281 ret = bdrv_pwrite(bs->file, off, to_write, (uint8_t *)s->header + off,
53fb7844 282 0);
0d31c7c2
DL
283 if (ret < 0) {
284 qemu_co_mutex_unlock(&s->lock);
285 return ret;
286 }
287 bit = find_next_bit(s->bat_dirty_bmap, size, bit + 1);
288 }
289 bitmap_zero(s->bat_dirty_bmap, size);
290
291 qemu_co_mutex_unlock(&s->lock);
292 return 0;
293}
294
295
8e0cf59d
EB
296static int coroutine_fn parallels_co_block_status(BlockDriverState *bs,
297 bool want_zero,
298 int64_t offset,
299 int64_t bytes,
300 int64_t *pnum,
301 int64_t *map,
302 BlockDriverState **file)
dd3bed16
RK
303{
304 BDRVParallelsState *s = bs->opaque;
8e0cf59d 305 int count;
dd3bed16 306
8e0cf59d 307 assert(QEMU_IS_ALIGNED(offset | bytes, BDRV_SECTOR_SIZE));
dd3bed16 308 qemu_co_mutex_lock(&s->lock);
8e0cf59d
EB
309 offset = block_status(s, offset >> BDRV_SECTOR_BITS,
310 bytes >> BDRV_SECTOR_BITS, &count);
dd3bed16
RK
311 qemu_co_mutex_unlock(&s->lock);
312
8e0cf59d 313 *pnum = count * BDRV_SECTOR_SIZE;
dd3bed16
RK
314 if (offset < 0) {
315 return 0;
316 }
317
8e0cf59d 318 *map = offset * BDRV_SECTOR_SIZE;
ddf4987d 319 *file = bs->file->bs;
8e0cf59d 320 return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID;
dd3bed16
RK
321}
322
5a41e1fa 323static coroutine_fn int parallels_co_writev(BlockDriverState *bs,
e18a58b4
EB
324 int64_t sector_num, int nb_sectors,
325 QEMUIOVector *qiov, int flags)
5a41e1fa
DL
326{
327 BDRVParallelsState *s = bs->opaque;
328 uint64_t bytes_done = 0;
329 QEMUIOVector hd_qiov;
330 int ret = 0;
331
e18a58b4 332 assert(!flags);
5a41e1fa
DL
333 qemu_iovec_init(&hd_qiov, qiov->niov);
334
335 while (nb_sectors > 0) {
336 int64_t position;
337 int n, nbytes;
338
339 qemu_co_mutex_lock(&s->lock);
ddd2ef2c 340 position = allocate_clusters(bs, sector_num, nb_sectors, &n);
5a41e1fa
DL
341 qemu_co_mutex_unlock(&s->lock);
342 if (position < 0) {
343 ret = (int)position;
344 break;
345 }
346
5a41e1fa
DL
347 nbytes = n << BDRV_SECTOR_BITS;
348
349 qemu_iovec_reset(&hd_qiov);
350 qemu_iovec_concat(&hd_qiov, qiov, bytes_done, nbytes);
351
d08c2a24
EB
352 ret = bdrv_co_pwritev(bs->file, position * BDRV_SECTOR_SIZE, nbytes,
353 &hd_qiov, 0);
5a41e1fa
DL
354 if (ret < 0) {
355 break;
356 }
357
358 nb_sectors -= n;
359 sector_num += n;
360 bytes_done += nbytes;
361 }
362
363 qemu_iovec_destroy(&hd_qiov);
364 return ret;
365}
366
481fb9cf
DL
367static coroutine_fn int parallels_co_readv(BlockDriverState *bs,
368 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov)
6ada7453 369{
29442569 370 BDRVParallelsState *s = bs->opaque;
481fb9cf
DL
371 uint64_t bytes_done = 0;
372 QEMUIOVector hd_qiov;
373 int ret = 0;
374
375 qemu_iovec_init(&hd_qiov, qiov->niov);
29442569 376
6ada7453 377 while (nb_sectors > 0) {
481fb9cf
DL
378 int64_t position;
379 int n, nbytes;
380
381 qemu_co_mutex_lock(&s->lock);
6953d920 382 position = block_status(s, sector_num, nb_sectors, &n);
481fb9cf
DL
383 qemu_co_mutex_unlock(&s->lock);
384
481fb9cf
DL
385 nbytes = n << BDRV_SECTOR_BITS;
386
bcbb3866
EK
387 qemu_iovec_reset(&hd_qiov);
388 qemu_iovec_concat(&hd_qiov, qiov, bytes_done, nbytes);
389
481fb9cf 390 if (position < 0) {
bcbb3866 391 if (bs->backing) {
d08c2a24
EB
392 ret = bdrv_co_preadv(bs->backing, sector_num * BDRV_SECTOR_SIZE,
393 nbytes, &hd_qiov, 0);
bcbb3866
EK
394 if (ret < 0) {
395 break;
396 }
397 } else {
398 qemu_iovec_memset(&hd_qiov, 0, 0, nbytes);
399 }
481fb9cf 400 } else {
d08c2a24
EB
401 ret = bdrv_co_preadv(bs->file, position * BDRV_SECTOR_SIZE, nbytes,
402 &hd_qiov, 0);
29442569 403 if (ret < 0) {
481fb9cf 404 break;
29442569 405 }
9d8b88f6 406 }
481fb9cf 407
9de9da17
RK
408 nb_sectors -= n;
409 sector_num += n;
481fb9cf 410 bytes_done += nbytes;
6ada7453 411 }
6ada7453 412
481fb9cf 413 qemu_iovec_destroy(&hd_qiov);
2914caa0
PB
414 return ret;
415}
416
49ad6467 417
2fd61638
PB
418static int coroutine_fn parallels_co_check(BlockDriverState *bs,
419 BdrvCheckResult *res,
420 BdrvCheckMode fix)
49ad6467
DL
421{
422 BDRVParallelsState *s = bs->opaque;
423 int64_t size, prev_off, high_off;
424 int ret;
425 uint32_t i;
426 bool flush_bat = false;
49ad6467 427
9a4f4c31 428 size = bdrv_getlength(bs->file->bs);
49ad6467
DL
429 if (size < 0) {
430 res->check_errors++;
431 return size;
432 }
433
2fd61638 434 qemu_co_mutex_lock(&s->lock);
6dd6b9f1
DL
435 if (s->header_unclean) {
436 fprintf(stderr, "%s image was not closed correctly\n",
437 fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR");
438 res->corruptions++;
439 if (fix & BDRV_FIX_ERRORS) {
440 /* parallels_close will do the job right */
441 res->corruptions_fixed++;
442 s->header_unclean = false;
443 }
444 }
445
49ad6467
DL
446 res->bfi.total_clusters = s->bat_size;
447 res->bfi.compressed_clusters = 0; /* compression is not supported */
448
449 high_off = 0;
450 prev_off = 0;
451 for (i = 0; i < s->bat_size; i++) {
452 int64_t off = bat2sect(s, i) << BDRV_SECTOR_BITS;
453 if (off == 0) {
454 prev_off = 0;
455 continue;
456 }
457
458 /* cluster outside the image */
459 if (off > size) {
460 fprintf(stderr, "%s cluster %u is outside image\n",
461 fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR", i);
462 res->corruptions++;
463 if (fix & BDRV_FIX_ERRORS) {
464 prev_off = 0;
465 s->bat_bitmap[i] = 0;
466 res->corruptions_fixed++;
467 flush_bat = true;
468 continue;
469 }
470 }
471
472 res->bfi.allocated_clusters++;
473 if (off > high_off) {
474 high_off = off;
475 }
476
e0b5207f 477 if (prev_off != 0 && (prev_off + s->cluster_size) != off) {
49ad6467
DL
478 res->bfi.fragmented_clusters++;
479 }
480 prev_off = off;
481 }
482
2fd61638 483 ret = 0;
49ad6467 484 if (flush_bat) {
a8f0e83c 485 ret = bdrv_co_pwrite_sync(bs->file, 0, s->header_size, s->header, 0);
49ad6467
DL
486 if (ret < 0) {
487 res->check_errors++;
2fd61638 488 goto out;
49ad6467
DL
489 }
490 }
491
e0b5207f 492 res->image_end_offset = high_off + s->cluster_size;
49ad6467
DL
493 if (size > res->image_end_offset) {
494 int64_t count;
e0b5207f 495 count = DIV_ROUND_UP(size - res->image_end_offset, s->cluster_size);
49ad6467
DL
496 fprintf(stderr, "%s space leaked at the end of the image %" PRId64 "\n",
497 fix & BDRV_FIX_LEAKS ? "Repairing" : "ERROR",
498 size - res->image_end_offset);
499 res->leaks += count;
500 if (fix & BDRV_FIX_LEAKS) {
ed3d2ec9 501 Error *local_err = NULL;
e8d04f92
HR
502
503 /*
504 * In order to really repair the image, we must shrink it.
505 * That means we have to pass exact=true.
506 */
507 ret = bdrv_truncate(bs->file, res->image_end_offset, true,
7b8e4857 508 PREALLOC_MODE_OFF, 0, &local_err);
49ad6467 509 if (ret < 0) {
ed3d2ec9 510 error_report_err(local_err);
49ad6467 511 res->check_errors++;
2fd61638 512 goto out;
49ad6467
DL
513 }
514 res->leaks_fixed += count;
515 }
516 }
517
2fd61638
PB
518out:
519 qemu_co_mutex_unlock(&s->lock);
520 return ret;
49ad6467
DL
521}
522
523
1511b490
KW
524static int coroutine_fn parallels_co_create(BlockdevCreateOptions* opts,
525 Error **errp)
74cf6c50 526{
1511b490
KW
527 BlockdevCreateOptionsParallels *parallels_opts;
528 BlockDriverState *bs;
529 BlockBackend *blk;
74cf6c50 530 int64_t total_size, cl_size;
369f7de9 531 uint32_t bat_entries, bat_sectors;
74cf6c50 532 ParallelsHeader header;
1511b490 533 uint8_t tmp[BDRV_SECTOR_SIZE];
74cf6c50
DL
534 int ret;
535
1511b490
KW
536 assert(opts->driver == BLOCKDEV_DRIVER_PARALLELS);
537 parallels_opts = &opts->u.parallels;
538
539 /* Sanity checks */
540 total_size = parallels_opts->size;
541
542 if (parallels_opts->has_cluster_size) {
543 cl_size = parallels_opts->cluster_size;
544 } else {
545 cl_size = DEFAULT_CLUSTER_SIZE;
546 }
547
2332d825
KW
548 /* XXX What is the real limit here? This is an insanely large maximum. */
549 if (cl_size >= INT64_MAX / MAX_PARALLELS_IMAGE_FACTOR) {
550 error_setg(errp, "Cluster size is too large");
551 return -EINVAL;
552 }
555a608c 553 if (total_size >= MAX_PARALLELS_IMAGE_FACTOR * cl_size) {
1511b490 554 error_setg(errp, "Image size is too large for this cluster size");
555a608c
KK
555 return -E2BIG;
556 }
74cf6c50 557
1511b490
KW
558 if (!QEMU_IS_ALIGNED(total_size, BDRV_SECTOR_SIZE)) {
559 error_setg(errp, "Image size must be a multiple of 512 bytes");
560 return -EINVAL;
74cf6c50
DL
561 }
562
1511b490
KW
563 if (!QEMU_IS_ALIGNED(cl_size, BDRV_SECTOR_SIZE)) {
564 error_setg(errp, "Cluster size must be a multiple of 512 bytes");
565 return -EINVAL;
566 }
567
568 /* Create BlockBackend to write to the image */
569 bs = bdrv_open_blockdev_ref(parallels_opts->file, errp);
570 if (bs == NULL) {
8942764f 571 return -EIO;
74cf6c50 572 }
8942764f 573
a3aeeab5
EB
574 blk = blk_new_with_bs(bs, BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL,
575 errp);
576 if (!blk) {
577 ret = -EPERM;
1511b490
KW
578 goto out;
579 }
580 blk_set_allow_write_beyond_eof(blk, true);
8942764f 581
1511b490 582 /* Create image format */
369f7de9 583 bat_entries = DIV_ROUND_UP(total_size, cl_size);
2d68e22e 584 bat_sectors = DIV_ROUND_UP(bat_entry_off(bat_entries), cl_size);
369f7de9 585 bat_sectors = (bat_sectors * cl_size) >> BDRV_SECTOR_BITS;
74cf6c50
DL
586
587 memset(&header, 0, sizeof(header));
588 memcpy(header.magic, HEADER_MAGIC2, sizeof(header.magic));
589 header.version = cpu_to_le32(HEADER_VERSION);
590 /* don't care much about geometry, it is not used on image level */
908b1c84
KK
591 header.heads = cpu_to_le32(HEADS_NUMBER);
592 header.cylinders = cpu_to_le32(total_size / BDRV_SECTOR_SIZE
593 / HEADS_NUMBER / SEC_IN_CYL);
74cf6c50 594 header.tracks = cpu_to_le32(cl_size >> BDRV_SECTOR_BITS);
369f7de9 595 header.bat_entries = cpu_to_le32(bat_entries);
74cf6c50 596 header.nb_sectors = cpu_to_le64(DIV_ROUND_UP(total_size, BDRV_SECTOR_SIZE));
369f7de9 597 header.data_off = cpu_to_le32(bat_sectors);
74cf6c50
DL
598
599 /* write all the data */
600 memset(tmp, 0, sizeof(tmp));
601 memcpy(tmp, &header, sizeof(header));
602
a9262f55 603 ret = blk_pwrite(blk, 0, BDRV_SECTOR_SIZE, tmp, 0);
74cf6c50
DL
604 if (ret < 0) {
605 goto exit;
606 }
1511b490 607 ret = blk_pwrite_zeroes(blk, BDRV_SECTOR_SIZE,
d004bd52 608 (bat_sectors - 1) << BDRV_SECTOR_BITS, 0);
74cf6c50
DL
609 if (ret < 0) {
610 goto exit;
611 }
74cf6c50 612
1511b490
KW
613 ret = 0;
614out:
615 blk_unref(blk);
616 bdrv_unref(bs);
74cf6c50
DL
617 return ret;
618
619exit:
620 error_setg_errno(errp, -ret, "Failed to create Parallels image");
1511b490
KW
621 goto out;
622}
623
b92902df
ML
624static int coroutine_fn parallels_co_create_opts(BlockDriver *drv,
625 const char *filename,
1511b490
KW
626 QemuOpts *opts,
627 Error **errp)
628{
629 BlockdevCreateOptions *create_options = NULL;
1511b490 630 BlockDriverState *bs = NULL;
92adf9db 631 QDict *qdict;
1511b490
KW
632 Visitor *v;
633 int ret;
634
635 static const QDictRenames opt_renames[] = {
636 { BLOCK_OPT_CLUSTER_SIZE, "cluster-size" },
637 { NULL, NULL },
638 };
639
640 /* Parse options and convert legacy syntax */
641 qdict = qemu_opts_to_qdict_filtered(opts, NULL, &parallels_create_opts,
642 true);
643
644 if (!qdict_rename_keys(qdict, opt_renames, errp)) {
645 ret = -EINVAL;
646 goto done;
647 }
648
649 /* Create and open the file (protocol layer) */
668f62ec 650 ret = bdrv_create_file(filename, opts, errp);
1511b490 651 if (ret < 0) {
1511b490
KW
652 goto done;
653 }
654
655 bs = bdrv_open(filename, NULL, NULL,
656 BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_PROTOCOL, errp);
657 if (bs == NULL) {
658 ret = -EIO;
659 goto done;
660 }
661
662 /* Now get the QAPI type BlockdevCreateOptions */
663 qdict_put_str(qdict, "driver", "parallels");
664 qdict_put_str(qdict, "file", bs->node_name);
665
af91062e
MA
666 v = qobject_input_visitor_new_flat_confused(qdict, errp);
667 if (!v) {
1511b490
KW
668 ret = -EINVAL;
669 goto done;
670 }
671
b11a093c 672 visit_type_BlockdevCreateOptions(v, NULL, &create_options, errp);
1511b490 673 visit_free(v);
b11a093c 674 if (!create_options) {
1511b490
KW
675 ret = -EINVAL;
676 goto done;
677 }
678
679 /* Silently round up sizes */
680 create_options->u.parallels.size =
681 ROUND_UP(create_options->u.parallels.size, BDRV_SECTOR_SIZE);
682 create_options->u.parallels.cluster_size =
683 ROUND_UP(create_options->u.parallels.cluster_size, BDRV_SECTOR_SIZE);
684
685 /* Create the Parallels image (format layer) */
686 ret = parallels_co_create(create_options, errp);
687 if (ret < 0) {
688 goto done;
689 }
690 ret = 0;
691
692done:
cb3e7f08 693 qobject_unref(qdict);
1511b490
KW
694 bdrv_unref(bs);
695 qapi_free_BlockdevCreateOptions(create_options);
696 return ret;
74cf6c50
DL
697}
698
23d6bd3b
DL
699
700static int parallels_probe(const uint8_t *buf, int buf_size,
701 const char *filename)
702{
703 const ParallelsHeader *ph = (const void *)buf;
704
705 if (buf_size < sizeof(ParallelsHeader)) {
706 return 0;
707 }
708
709 if ((!memcmp(ph->magic, HEADER_MAGIC, 16) ||
710 !memcmp(ph->magic, HEADER_MAGIC2, 16)) &&
711 (le32_to_cpu(ph->version) == HEADER_VERSION)) {
712 return 100;
713 }
714
715 return 0;
716}
717
6dd6b9f1
DL
718static int parallels_update_header(BlockDriverState *bs)
719{
720 BDRVParallelsState *s = bs->opaque;
9a4f4c31
KW
721 unsigned size = MAX(bdrv_opt_mem_align(bs->file->bs),
722 sizeof(ParallelsHeader));
6dd6b9f1
DL
723
724 if (size > s->header_size) {
725 size = s->header_size;
726 }
32cc71de 727 return bdrv_pwrite_sync(bs->file, 0, size, s->header, 0);
6dd6b9f1
DL
728}
729
23d6bd3b
DL
730static int parallels_open(BlockDriverState *bs, QDict *options, int flags,
731 Error **errp)
732{
733 BDRVParallelsState *s = bs->opaque;
734 ParallelsHeader ph;
19f5dc15 735 int ret, size, i;
d6179011
DL
736 QemuOpts *opts = NULL;
737 Error *local_err = NULL;
738 char *buf;
23d6bd3b 739
83930780
VSO
740 ret = bdrv_open_file_child(NULL, options, "file", bs, errp);
741 if (ret < 0) {
742 return ret;
4e4bf5c4
KW
743 }
744
32cc71de 745 ret = bdrv_pread(bs->file, 0, sizeof(ph), &ph, 0);
23d6bd3b
DL
746 if (ret < 0) {
747 goto fail;
748 }
749
750 bs->total_sectors = le64_to_cpu(ph.nb_sectors);
751
752 if (le32_to_cpu(ph.version) != HEADER_VERSION) {
753 goto fail_format;
754 }
755 if (!memcmp(ph.magic, HEADER_MAGIC, 16)) {
756 s->off_multiplier = 1;
757 bs->total_sectors = 0xffffffff & bs->total_sectors;
758 } else if (!memcmp(ph.magic, HEADER_MAGIC2, 16)) {
759 s->off_multiplier = le32_to_cpu(ph.tracks);
760 } else {
761 goto fail_format;
762 }
763
764 s->tracks = le32_to_cpu(ph.tracks);
765 if (s->tracks == 0) {
766 error_setg(errp, "Invalid image: Zero sectors per track");
767 ret = -EINVAL;
768 goto fail;
769 }
770 if (s->tracks > INT32_MAX/513) {
771 error_setg(errp, "Invalid image: Too big cluster");
772 ret = -EFBIG;
773 goto fail;
774 }
e0b5207f 775 s->cluster_size = s->tracks << BDRV_SECTOR_BITS;
23d6bd3b
DL
776
777 s->bat_size = le32_to_cpu(ph.bat_entries);
778 if (s->bat_size > INT_MAX / sizeof(uint32_t)) {
779 error_setg(errp, "Catalog too large");
780 ret = -EFBIG;
781 goto fail;
782 }
783
2d68e22e 784 size = bat_entry_off(s->bat_size);
9a4f4c31
KW
785 s->header_size = ROUND_UP(size, bdrv_opt_mem_align(bs->file->bs));
786 s->header = qemu_try_blockalign(bs->file->bs, s->header_size);
23d6bd3b
DL
787 if (s->header == NULL) {
788 ret = -ENOMEM;
789 goto fail;
790 }
19f5dc15
DL
791 s->data_end = le32_to_cpu(ph.data_off);
792 if (s->data_end == 0) {
793 s->data_end = ROUND_UP(bat_entry_off(s->bat_size), BDRV_SECTOR_SIZE);
794 }
795 if (s->data_end < s->header_size) {
23d6bd3b
DL
796 /* there is not enough unused space to fit to block align between BAT
797 and actual data. We can't avoid read-modify-write... */
798 s->header_size = size;
799 }
800
32cc71de 801 ret = bdrv_pread(bs->file, 0, s->header_size, s->header, 0);
23d6bd3b
DL
802 if (ret < 0) {
803 goto fail;
804 }
805 s->bat_bitmap = (uint32_t *)(s->header + 1);
806
19f5dc15
DL
807 for (i = 0; i < s->bat_size; i++) {
808 int64_t off = bat2sect(s, i);
809 if (off >= s->data_end) {
810 s->data_end = off + s->tracks;
811 }
812 }
813
6dd6b9f1
DL
814 if (le32_to_cpu(ph.inuse) == HEADER_INUSE_MAGIC) {
815 /* Image was not closed correctly. The check is mandatory */
816 s->header_unclean = true;
817 if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) {
818 error_setg(errp, "parallels: Image was not closed correctly; "
819 "cannot be opened read/write");
820 ret = -EACCES;
821 goto fail;
822 }
823 }
824
c6ecec43
MA
825 opts = qemu_opts_create(&parallels_runtime_opts, NULL, 0, errp);
826 if (!opts) {
d6179011
DL
827 goto fail_options;
828 }
829
a5f9b9df 830 if (!qemu_opts_absorb_qdict(opts, options, errp)) {
d6179011
DL
831 goto fail_options;
832 }
833
834 s->prealloc_size =
835 qemu_opt_get_size_del(opts, PARALLELS_OPT_PREALLOC_SIZE, 0);
836 s->prealloc_size = MAX(s->tracks, s->prealloc_size >> BDRV_SECTOR_BITS);
837 buf = qemu_opt_get_del(opts, PARALLELS_OPT_PREALLOC_MODE);
bda4cdcb 838 /* prealloc_mode can be downgraded later during allocate_clusters */
f7abe0ec 839 s->prealloc_mode = qapi_enum_parse(&prealloc_mode_lookup, buf,
06c60b6c
MA
840 PRL_PREALLOC_MODE_FALLOCATE,
841 &local_err);
d6179011
DL
842 g_free(buf);
843 if (local_err != NULL) {
4bc6d7ee 844 error_propagate(errp, local_err);
d6179011
DL
845 goto fail_options;
846 }
dc62da88 847
baefd977
VSO
848 if (ph.ext_off) {
849 if (flags & BDRV_O_RDWR) {
850 /*
851 * It's unsafe to open image RW if there is an extension (as we
852 * don't support it). But parallels driver in QEMU historically
853 * ignores the extension, so print warning and don't care.
854 */
855 warn_report("Format Extension ignored in RW mode");
856 } else {
857 ret = parallels_read_format_extension(
858 bs, le64_to_cpu(ph.ext_off) << BDRV_SECTOR_BITS, errp);
859 if (ret < 0) {
860 goto fail;
861 }
862 }
863 }
864
6c7d390b 865 if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_INACTIVE)) {
6dd6b9f1
DL
866 s->header->inuse = cpu_to_le32(HEADER_INUSE_MAGIC);
867 ret = parallels_update_header(bs);
868 if (ret < 0) {
869 goto fail;
870 }
871 }
872
8e3b0cbb 873 s->bat_dirty_block = 4 * qemu_real_host_page_size();
0d31c7c2
DL
874 s->bat_dirty_bmap =
875 bitmap_new(DIV_ROUND_UP(s->header_size, s->bat_dirty_block));
876
a94750d9 877 /* Disable migration until bdrv_activate method is added */
1d0f37cf
JC
878 error_setg(&s->migration_blocker, "The Parallels format used by node '%s' "
879 "does not support live migration",
880 bdrv_get_device_or_node_name(bs));
386f6c07
MA
881 ret = migrate_add_blocker(s->migration_blocker, errp);
882 if (ret < 0) {
1d0f37cf
JC
883 error_free(s->migration_blocker);
884 goto fail;
885 }
23d6bd3b
DL
886 qemu_co_mutex_init(&s->lock);
887 return 0;
888
889fail_format:
890 error_setg(errp, "Image not in Parallels format");
4bc6d7ee 891fail_options:
23d6bd3b
DL
892 ret = -EINVAL;
893fail:
894 qemu_vfree(s->header);
895 return ret;
896}
897
898
6ada7453
TS
899static void parallels_close(BlockDriverState *bs)
900{
901 BDRVParallelsState *s = bs->opaque;
6dd6b9f1 902
6c7d390b 903 if ((bs->open_flags & BDRV_O_RDWR) && !(bs->open_flags & BDRV_O_INACTIVE)) {
6dd6b9f1
DL
904 s->header->inuse = 0;
905 parallels_update_header(bs);
e8d04f92
HR
906
907 /* errors are ignored, so we might as well pass exact=true */
908 bdrv_truncate(bs->file, s->data_end << BDRV_SECTOR_BITS, true,
7b8e4857 909 PREALLOC_MODE_OFF, 0, NULL);
19f5dc15
DL
910 }
911
0d31c7c2 912 g_free(s->bat_dirty_bmap);
9eae9cca 913 qemu_vfree(s->header);
1d0f37cf
JC
914
915 migrate_del_blocker(s->migration_blocker);
916 error_free(s->migration_blocker);
6ada7453
TS
917}
918
5efa9d5a 919static BlockDriver bdrv_parallels = {
e60f469c
AJ
920 .format_name = "parallels",
921 .instance_size = sizeof(BDRVParallelsState),
922 .bdrv_probe = parallels_probe,
1dec5a70 923 .bdrv_open = parallels_open,
e60f469c 924 .bdrv_close = parallels_close,
69dca43d 925 .bdrv_child_perm = bdrv_default_perms,
8e0cf59d 926 .bdrv_co_block_status = parallels_co_block_status,
d0e61ce5 927 .bdrv_has_zero_init = bdrv_has_zero_init_1,
0d31c7c2 928 .bdrv_co_flush_to_os = parallels_co_flush_to_os,
481fb9cf 929 .bdrv_co_readv = parallels_co_readv,
5a41e1fa 930 .bdrv_co_writev = parallels_co_writev,
d67066d8 931 .is_format = true,
bcbb3866 932 .supports_backing = true,
1511b490 933 .bdrv_co_create = parallels_co_create,
efc75e2a 934 .bdrv_co_create_opts = parallels_co_create_opts,
2fd61638 935 .bdrv_co_check = parallels_co_check,
74cf6c50 936 .create_opts = &parallels_create_opts,
6ada7453 937};
5efa9d5a
AL
938
939static void bdrv_parallels_init(void)
940{
941 bdrv_register(&bdrv_parallels);
942}
943
944block_init(bdrv_parallels_init);